Академический Документы
Профессиональный Документы
Культура Документы
PROCESS
Liverpool John Moores University, Control Systems Research Group, School of Engineering, Byrom Street,
Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 16,2010 at 04:03:56 EDT from IEEE Xplore. Restrictions apply.
(1 993) and a realistic estimation of the suspended solids CONTROLLERS
concentration found in the effluent (Dupont & Henze,
1992). The multi-input, multi-output activated sludge
model, has been successfully implemented within Three controllers have been investigated for aeration
MATLAB. The system simulates control of aeration in control: a simple PID controller since it is an industry
an aeration lane represented by 3 tanks (Figure I), and standard, widely available and is the most likely system
the regulation of the concentrations of solids in the to be implemented on a real plant. A Fuzzy logic
aeration and clarifier tanks. controller has also been developed owing to its
increasing interest in industry, PLC’s with fuzzy
The dissolved oxygen ‘reading’ is filtered with a first capabilities are available even if not yet widely used.
order transfer function simulating an industry standard Finally, a self-tuning coIltroller was implemented
dissolved oxygen probe, with a 3 minute time constant. because it is a technique that is likely to have an impact
Also, a normally distributed white noise of 1% is added in the near future to accommodate non-linearity and time
to the measurement. variance.
The input data used represents a typical week with The flow control consists of a PID controller with a very
weekly and diurnal flow and concentration patterns slow time constant since variations in recycled sludge
(Figures 2 and 3). It is based on real data (Spanjers et flow are a major disturbance to the aeration, and a
al, 1997) and incorporates two storm events (day 1 and second PID for surplus sludge flow control.
day 4). The average flow is 20 000 m3d-’and total COD
of 300 mgl-’ for ii population equivalent of 100 000. The control of aeration, in the 2 aerobic tanks, is
The two storms produce different effects even though, in independent fiom each other, according to the dissolved
both cases, the flow increases 3 fold to 60 000 m3day-’. oxygen measurement given by a DO probe placed in
In the first storm, ithe soluble substrate decreases by 15 each tank. The sample time used was 2 minutes.
% while the particulate substrate increases 4 folds.
During the second storm event, the soluble substrate is A standard PID controller is used. Its main drawback is
decreased by 15% but the particulate substrate is the fine-tuning which is difficult. A rough tuning can be
reduced to half its thy weather value simulating the wash achieved rapidly by standard methods such as Ziegler-
out that happened during the first storm. Nichols, but fine-tuning is a long and delicate operation,
MluentFbw difficult to perform on an on-line system.
RESULTS
1103
Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 16,2010 at 04:03:56 EDT from IEEE Xplore. Restrictions apply.
controller presents a maximum error close to 3.5% while CONCLUSION
the self-tuning controller error is less than 2.5%.
The STC presents the least variation of the dissolved To summarise, it can be said that the quickest way to
oxygen concentration (Figure 8). The PID and hzzy implement a control system for dissolved oxygen
logic control (Figure 4 and 6 respectively) offer quite concentration in an activated sludge process would be to
similar performances except during the fist storm event use a self-tuning controller. Overall it presents a good
where the PID responds slightly better than the FLC. response to a variety of conditions and offers the
The control signals (i.e. the aeration rate) are also very advantage of a certain degree of robustness over the
similar. However, the aeration rate delivered by the changes that the process can undergo. A PID controller
STC is noisier: the controller tries to correct the can do nearly as well, and has the advantage of
measurement noise (it disappears if the noise is simplicity but does present the problem of parameter
removed). This requires further investigation, since fast tuning and the lack of adaptability to naturally occurring
variations of the aeration rate, have the undesirable process changes. Overall, the fuzzy logic controller
effect of excessive valves movements. The PID and offers performances between the other two controllers
FLC seem less prone to this behaviour. for both complexity and robustness. In particular, the
noise rejection ability is better than for the PID
controller, however, the main problem with the FLC is
O p t i ~ i § a t ~ oofnactivated sludge flows and costs rule-base generation.
1104
Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 16,2010 at 04:03:56 EDT from IEEE Xplore. Restrictions apply.
DO (PID controller)
2.00 k ~. A 0.10
2.00 -4~
" "" .I._"
I I 0.10
- 1 . 5 0 1 ,!I ; 0.05 =
I
a
5 - 1.50 f -0.05B
-P
2 0.00
B
E
f
81 l.OO
8
-100
8
ow
g
0.50 -0.05
5ror
ow -0 10 0.00 1 -0 10
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Time (drys) Time (days)
'" I I
- I
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Time (days) Time (days)
Figure 5: Aeration rate with PID controller Figure 9: Aeration rate with self-tuning controller
1
Comparison of optimisation effects
8w 20
15 3
0
Ti
V
-
I
; 400 ,ogs
gg
1 i!
3 200 5 %
cn
0.w D 1 I -0.10 0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Tim. (dry@ Tima (d.vr)
60
250
5 40
30
20
0 1
0 1 2 3 4 5 6 7
Time (days)
1105
Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 16,2010 at 04:03:56 EDT from IEEE Xplore. Restrictions apply.